With the wide use of multiprocessor architecture, parallel tasks have been receiving growing attention in both industry and academia. In real-time systems, the scheduling and synchronization that ensure predictable task execution and resource access are of utmost importance. Although the scheduling of (independent) parallel tasks is widely studied in recent years, few works have been done for the synchronization with intra-task parallelism. In particular, the performance of the classical priority inheritance mechanism is still unfathomed for parallel tasks. In this paper, we extend the Priority Inheritance Protocol (PIP) and the Flexible Multiprocessor Locking Protocol (FMLP) for real-time parallel tasks (i.e., called P-PIP and P-FMLP respectively) under the global fixed-priority scheduling, and present the blocking analysis for both protocols with the state-of-the-art linear optimization technique. Empirical evaluations show that the P-PIP outperformed the P-FMLP with increasing shared resource contentions.

随着多处理器体系结构的广泛使用，并行任务在业界和学术界都越来越受到关注。在实时系统中，确保可预测的任务执行和资源访问的调度和同步至关重要。尽管近年来对（独立）并行任务的调度进行了广泛的研究，但与任务内并行进行同步的工作很少。尤其是，对于并行任务，经典优先级继承机制的性能仍然无法确定。在本文中，我们针对全局固定优先级调度下的实时并行任务（即分别称为P-PIP和P-FMLP）扩展了优先级继承协议（PIP）和灵活多处理器锁定协议（FMLP），并使用最新的线性优化技术对这两种协议进行了阻塞分析。实证评估表明，随着共享资源争用的增加，P-PIP的性能优于P-FMLP。